Don't get rugged
the score first.
Inspect any Solana coin in seconds. Find out if it's bundled, a potential rug, or a pure gem — with the on-chain evidence behind every call. No wallet. No signing.
- Mint authorities
- Pools & liquidity
- Metadata & registries
- Holder distribution
- LP lock status
- Honeypot simulation
- Bundle & snipers
- Holder count
- Mint & freeze authority
- LP burned or pullable
- Bundle & sniper detection
- Holder concentration
- Honeypot simulation
- Metadata & socials
Not financial advice. A high CLEAN SCORE means the structural rug vectors we test are closed — it cannot tell you the team won't sell, that the chart won't die, or that an off-chain promise will be kept. Soft rugs need no permissions. Always size accordingly.
Agent Check
A memecoin with a receipt. The scanner above is the utility, it is free forever, and it runs entirely in your browser. $CHECK is the ticker that pays for it.
Agent Check
$CHECK · Solana · pump.fun
Launching soon — CA drops here on release
The CA appears here the moment the token launches. Until then, every address claiming to be $CHECK is fake — verify against this page only.
Why a token at all?
- The scanner is free, unlimited and needs no wallet — the token funds the RPC bill, not a paywall.
- Holders get the deep-scan endpoints: full holder counts, bundle detection and wallet clustering.
- Every feature ships publicly first. No promises held hostage behind a buy button.
- We scan our own contract, on our own site, and show the score whether it flatters us or not.
Roadmap
- Phase 1 — LivePublic scanner, 16 criteria, evidence-backed reports, share links.
- Phase 2 — LiveHoneypot round-trip simulation, LP lock resolution, circulating-adjusted concentration.
- Phase 3 — LiveLaunch forensics: bundle detection, sniper detection, and live overhang tracking of what those wallets still hold.
- Phase 4Wallet clustering and deployer history — "has this wallet rugged before?"
- Phase 5Telegram & X scan bot, watchlists, live alerts when an LP moves or an authority changes.
- Phase 6Public API + embeddable score badge so any site can show a $CHECK score.
What we actually check
Most "rug checkers" show you two booleans and call it a day. $CHECK runs 19 weighted criteria across six categories, applies 16 hard ceilings for the failures that make everything else irrelevant, and shows you the evidence for every single one.
01 Weighted, not checklist
Each criterion carries a weight reflecting how often it is the actual mechanism of a rug. Mint authority is worth ten times a missing Telegram link, because it is.
02 Unknowns are excluded, not guessed
If an RPC can't answer, that check leaves the denominator entirely and your confidence figure drops. We never invent a passing grade from missing data.
03 Hard ceilings override the average
Live freeze authority caps the score at 40. A honeypot caps it at 12. No amount of liquidity, volume or blue checkmarks can lift a token past a ceiling.
04 Circulating-adjusted concentration
We resolve every top holder's owner, identify which are AMM vaults, lockers or burn addresses, and measure concentration against the float that can actually be sold.
How to read the score
FAQ
Do I need to connect a wallet?
No — and you never will. The scanner only makes read calls to public Solana RPC nodes and market APIs. There is no connect button on this site by design, because a rug checker that asks you to sign something is the joke that writes itself.
Can a 100/100 token still go to zero?
Absolutely. The score measures structural rug vectors: can supply be inflated, can you be frozen, can the pool be pulled, can you actually sell. It cannot measure intent. A team with a clean contract can still stop showing up — that's a soft rug and no on-chain scanner can see it coming.
How does bundle and sniper detection work?
We page back through the token's entire signature history to its very first transactions, then replay the token-balance changes block by block. Wallets that bought in the same block trading opened are a bundle — on pump.fun that means one Jito bundle containing the deployer's create, their own buy, and a fleet of wallets nobody could outbid. Wallets that got in over the next few seconds are snipers. We then read what each of those wallets holds today, so you can tell a bundle that already dumped from one still sitting on your exit liquidity.
This is by far the most expensive check we run, because its cost scales with how much the token has traded — so on the shared public RPC it does not run automatically. Every report has a Run bundle & sniper scan button that starts it on demand. Set a private RPC in settings (the gear icon) and it runs as part of every scan instead.
If it cannot read enough of the launch, it says so rather than showing you a number. A partial replay always understates a bundle, and a checker that reports "3% bundled" when the truth is 58% is worse than one that admits it does not know.
Why does my token show "unknown" on some checks?
Public RPC endpoints rate-limit hard and some refuse expensive calls entirely. Rather than guessing, we drop unverifiable checks from the score and lower the confidence figure instead. Paste a private RPC URL in settings (the gear icon) and most unknowns resolve instantly.
How does the honeypot test work without my money?
We ask Jupiter to build — not send — a route that buys 1 SOL worth of the token, then a second route selling 100% of that result back to SOL. Nothing is broadcast and no signature exists. If the buy routes but the sell doesn't, or the round trip loses a punitive share, that shows up immediately.
What's the difference between the score and the ceiling?
The score is a weighted average of everything we could verify. The ceiling is a hard cap applied when a single failure makes the rest irrelevant — a live freeze authority caps you at 40 no matter how beautiful the chart is. When a ceiling is active we tell you which one and why.
Is this open source?
Yes. Every criterion, weight and threshold lives in js/core/scoring.js in plain readable JavaScript. You can audit the maths, disagree with a weight, and change it. A scoring engine you can't inspect is just vibes with extra steps.